Improving Stability Estimates in Adversarial Explainable AI through Alternate Search Methods